Yukang-Lian opened a new pull request, #67178:
URL: https://github.com/apache/doris/pull/67178
### What problem does this PR solve?
Issue Number: None
Related PR: #49383
Problem Summary:
Size-based cumulative compaction currently promotes a merge-on-write output
rowset to the base layer once its version span exceeds
`compaction_promotion_version_count`, even when the output rowset is only a few
hundred KB. For high-frequency tiny loads, this moves small rowsets into the
base layer prematurely and causes frequent base compactions.
PR #49383 makes cumulative compaction aggregate historical delete bitmaps
into the output rowset version and remove the pre-rowset delete bitmap
versions. With that optimization enabled, raw rowset version span is no longer
needed as a proxy for delete bitmap pressure.
This PR removes version-count promotion from both local and cloud size-based
cumulative compaction policies, removes the obsolete configuration, and keeps
size-based and delete-version promotion unchanged.
### Release note
MOW cumulative compaction no longer promotes small output rowsets based only
on version count.
### Check List (For Author)
- Test
- [ ] Regression test
- [x] Unit Test
- [ ] Manual test (add detailed scripts or steps below)
- [ ] No need to test or manual test. Explain why:
- [ ] This is a refactor/code format and no logic has been changed.
- [ ] Previous test can cover this change.
- [ ] No code files have been changed.
- [ ] Other reason
Focused ASAN BE UT: 68/68 passed for
`TestSizeBasedCumulativeCompactionPolicy.*` and
`TestCloudSizeBasedCumulativeCompactionPolicy.*`.
- Behavior changed:
- [ ] No.
- [x] Yes. Small MOW cumulative output rowsets remain in the cumulative
layer until the normal size or delete-version promotion condition is met.
- Does this need documentation?
- [x] No.
- [ ] Yes.
### QA validation
Please validate both local and cloud modes:
1. Create an MOW table and generate more than 1,000 tiny versions, then run
cumulative compaction. When the output remains below the size threshold and has
no delete version, verify that the cumulative point is not promoted solely by
version span and base compaction is not triggered repeatedly.
2. Verify that delete bitmap aggregation and removal introduced by #49383
keeps historical delete bitmap count bounded and query results correct.
3. Verify that reaching the size threshold or compacting a delete version
still promotes the cumulative point.
### Check List (For Reviewer who merge this PR)
- [ ] Confirm the release note
- [ ] Confirm test cases
- [ ] Confirm document
- [ ] Add branch pick label
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
